Noninvasive evaluation of the regional variations of GABA using magnetic resonance spectroscopy at 3 Tesla

Magnetic Resonance Imaging 33 (2015) 611–617

Contents lists available at ScienceDirect

Magnetic Resonance Imaging journal homepage: www.mrijournal.com

Noninvasive evaluation of the regional variations of GABA using magnetic resonance spectroscopy at 3 Tesla☆,☆☆ Christopher R. Durst a,⁎, 1, Navin Michael b, 1, Nicholas J. Tustison a, James T. Patrie c, Prashant Raghavan d, Max Wintermark a, S. Sendhil Velan b, e, f a

Department of Radiology and Medical Imaging, University of Virginia, Charlottesville, VA, USA Singapore Institute for Clinical Sciences, Agency for Science, Technology and Research, Singapore Department of Public Health Sciences, University of Virginia, Charlottesville, VA, USA d Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland, Baltimore, MD, USA e Laboratory of Molecular Imaging, Singapore Bioimaging Consortium, Agency for Science, Technology and Research, Singapore, Singapore f Clinical Imaging Research Centre, Agency for Science, Technology and Research, Singapore, NUS-A*STAR, Singapore b c

a r t i c l e

i n f o

Article history: Received 11 May 2014 Revised 29 August 2014 Accepted 16 February 2015 Keywords: GABA Spectroscopy MRS Segmentation Regional variation

a b s t r a c t Purpose: Rapid regional fluctuations in GABA may result in inhomogeneous concentrations throughout the brain parenchyma. The goal of this study is to provide further insight into the natural distribution of GABA throughout the brain and thus determine if a surrogate site may be used for spectroscopy when evaluating motor diseases, neurological disorders, or psychiatric dysfunction. Materials and Methods: In this prospective study, eight healthy volunteers underwent spectroscopic evaluation of the frontal lobe, occipital lobe, lateral temporal lobe, basal ganglia, and both hippocampi using a spin echo variant of a J-difference editing method. Knowledge of the relative peak intensities of the macromolecule peaks at 3 ppm and 0.9 ppm was used to correct the contribution of co-edited macromolecules to the GABA peak at 3 ppm. The GABA values were internally referenced to NAA. Linear regression was used to normalize the effect of regional tissue-fraction variation on the GABA/NAA values. A one-way ANOVA was performed with Tukey's multiple comparison test to compare the normalized GABA/ NAA values in each pair of locations. Results: After accounting for the macromolecule contribution to the GABA signal and correction for tissue fraction variation, the normalized GABA/NAA ratios differ significantly between the six brain locations (p b 0.001). Pairwise comparisons of the corrected normalized GABA/NAA ratios show statistically significant variation between the frontal lobe and the basal ganglia, frontal and lateral temporal lobes, and frontal lobe and right hippocampus. Variations in the normalized GABA/NAA ratios trend toward significance between the frontal lobe and left hippocampus, occipital lobe and the frontal lobe, occipital lobe and basal ganglia, and occipital lobe and right hippocampus. Conclusion: Our study suggests that GABA concentration is inhomogeneous throughout the parenchyma. Studies evaluating the role of GABA must carefully consider voxel placement when incorporating spectroscopy. © 2015 Elsevier Inc. All rights reserved.

1. Introduction As the principal inhibitory neurotransmitter, gamma aminobutyric acid (GABA) and GABAergic systems counterbalance the excitatory activity of glutametergic systems to regulate the firing

☆ Grant support: This project was generously funded by the Virginia Chapter of the American College of Radiology. ☆☆ Acknowledgments: None. ⁎ Corresponding author at: University of Virginia Health System, PO Box 800170, Charlottesville, VA 22908. Tel.: +1 434 982 1736; fax: +1 434 982 3880. E-mail address: [email protected] (C.R. Durst). 1 Contributed equally and would like to be considered co-first authors. http://dx.doi.org/10.1016/j.mri.2015.02.015 0730-725X/© 2015 Elsevier Inc. All rights reserved.

rate of neurons [1]. An imbalance in the system may result in a number of neurologic and psychiatric disorders, including epilepsy [2,3], depression [4,5], schizophrenia [6], autoimmune inflammation [7], and motor learning [8,9]. Control of these systems requires rapid, regionally specific modulation of neurotransmitter concentration [9]. Recent fMRI work suggests that BOLD responses may be at least partially attributable to these local alterations in GABA concentration [10,11]. However, these rapid regional fluctuations may result in inhomogeneous GABA concentrations throughout the parenchyma [12,13]. In this study, we provide further insight into the natural distribution of GABA throughout the brain in normal volunteers.

612

C.R. Durst et al. / Magnetic Resonance Imaging 33 (2015) 611–617

2. Materials and methods 2.1. Regulatory oversight This study was performed under the auspices of our institutional review board. Informed consent was obtained from each volunteer. The volunteers' personal information was protected according to HIPAA regulations. 2.2. Volunteer selection Due to reported effects of gender on GABA levels, only male volunteers were enrolled in the study [14]. The age range was restricted to volunteers between the ages of 28 and 40 years due to reported fluctuations in GABA levels with age [15]. Volunteers were screened for any medical issues, including (but not limited to) a history of seizures, neuropathic pain syndromes, movement disorders, and psychiatric disorders. 2.3. Study plan An MPRAGE study of the whole brain was initially obtained for planning and tissue segmentation. Next, each volunteer underwent spectroscopic interrogation of the left frontal lobe, the occipital lobe, the basal ganglia, the lateral left temporal lobe, and both hippocampi. The voxel locations were used to generate regions of interest (ROIs) for parenchymal segmentation for each study. The spectroscopy data were post-processed off-site. The data were then collated and evaluated by a blinded statistician. 2.4. MRI imaging and spectroscopy Imaging was performed on a Siemens Tim Trio 3 T scanner (IDEA VB 17) using an 8 channel receive only head coil (Siemens AG, Munich, Germany). A sagittal MPRAGE (TR/TE/TI, 1900/900/2.34) was initially performed for localization and study planning. Each patient then underwent a series of studies using a MEGA-PRESS single voxel spectroscopy sequence from Siemens Healthcare (Siemens Healthcare, Cary, North Carolina, USA). The sequence is a spin echo variant of the J-difference editing method originally developed by Mescher [16,17]. It uses a point resolved spectroscopy (PRESS) sequence with a MEGA suppression scheme. Parameters for MEGA-PRESS sequence were TE = 68 ms, TR = 1500 ms, bandwidth of 1200 Hz, 384 averages, vector size of 512, and a delta frequency of − 1.7 ppm. During odd-numbered acquisitions, a frequency selective refocusing Gaussian pulse was issued at 1.9 ppm. During the even-numbered acquisitions, the Gaussian pulse was irradiated symmetrically on the other side of the water peak at 7.5 ppm. The J-difference editing approach exploits the fact that the GABA spins at 3 ppm are coupled to the spins at 1.9 ppm, whereas the overlapping creatine signal at 3 ppm is not. The Gaussian pulse at 1.9 ppm in the odd-numbered acquisitions induces a phase reversal in the outer two peaks so that they refocus with opposite phase relative to the central, and more importantly, the non-coupled creatine peak. The Gaussian pulse at 7.5 ppm in the even acquisitions does not have any refocusing effect and allows free evolution of J-coupled spins. The creatine peak at 3 ppm is unaffected by the editing pulses and is removed in the difference spectra, while the outer two peaks of the GABA triplet are retained. Suppression of the water signal was achieved by chemical shift selective pulses (CHESS). Both automatic and manual shimming was performed to improve the homogeneity. Achieving a good signal to noise ratio (SNR) for quantitating GABA using the MEGA-PRESS SVS sequence requires a relatively large voxel due to the low concentration of GABA. The voxel size for

the acquisitions was fixed at 20 × 30 × 40 mm 3, regardless of location. Regions of interest included the frontal lobe, occipital lobe, bilateral hippocampi, lateral temporal lobe, and the basal ganglia. The acquisition time in each location was 9.6 min with a total acquisition time of 57.6 min. The frontal lobe, occipital lobe, lateral temporal lobe, and hippocampi were selected due to their inclusion in previous studies [18–21]. The hippocampi and lateral temporal lobe were also selected based on the correlation between GABA levels and epilepsy. The basal ganglia was selected due to its more uniform nature and greater density of grey matter. The frontal lobe voxel was oriented parallel to the frontal pole and positioned as anteriorly as possible while minimizing involvement of the caudate nucleus or corpus callosum. To match the orientation of previous studies, voxels in the occipital lobe spanned midline and were oriented parallel to the overlying occipital bone above the tentorium. When imaging the lateral temporal lobe, the voxel was positioned with the long axis matching the long axis of the temporal lobe and paralleling the overlying temporal bone. Voxels centered in the basal ganglia were positioned with the posterior margin of the voxel roughly parallel with the atrium of the lateral ventricle. The voxel was oriented at a slight transverse angle to avoid inclusion of the frontal horn while maximizing coverage of the thalamus, lentiform nucleus, and a portion of the caudate nucleus. For the hippocampi, axial and coronal images of the brain were reconstructed from the sagittal MPRAGE so that they would be parallel and perpendicular to the hippocampi, respectively. The voxel was oriented such that the long axis was parallel to the hippocampal formation with the anterior edge of the voxel at the posterior margin of the amygdala. After initial voxel placement, the technologist had the liberty to reorient the voxel to minimize the inclusion of CSF and increase the distance between the voxel and the skull base or calvarium. 2.5. Post-processing of spectroscopy data Metabolite quantitation of the MEGA-PRESS difference spectra was performed using TARQUIN [22]. TARQUIN is a time domain fitting tool that uses non-negative least squares projection to estimate the signal amplitudes in both in-vivo and ex-vivo 1H MRS data. The preprocessing steps included exponential line-broadening by 4 Hz, zero-filling by a factor of 4, automated phase-correction, removal of any residual water peak and baseline correction. The pseudo-doublet structure of the GABA signal in the difference spectrum was modeled as a sum of two Gaussian singlets. The macromolecule peak which resonates at 3 ppm (MM30) is coupled to the macromolecule peak at 1.7 ppm (MM17). Hence, the editing pulse applied at 1.9 ppm results in contamination of the GABA peak by co-edited macromolecules. The combined peak has been referred to as GABA+. The regional variation of GABA + may not reflect the true variation of GABA due to the heterogeneity of macromolecules. Some methods have been proposed for suppressing the co-edited macromolecules. With a symmetrical-editing approach, the editing pulse is mirrored at 1.5 ppm in the odd acquisitions [23]. Since the effect of the editing pulses on MM17 is identical in the even and odd acquisitions, the co-editing of the macromolecules can be avoided in the difference spectra. However, the limited selectivity of the editing pulses at the typical clinical field strengths of 1.5 T and 3 T can result in unintended suppression of the GABA signal. The usage of a longer echo time of 80 ms was suggested as a strategy for allowing longer and more selective editing pulses [24]. However, an echo time of 68 ms continues to be more widely used for MEGA-PRESS based detection of GABA. Other methods include the pre-inversion approach, in which a separate macromolecule spectrum is obtained by exploiting the differences in the T1 of metabolites and macromolecules, and subtracting this signal from the MEGA-PRESS difference spectrum [25]. However, the T1 of GABA has not been well

C.R. Durst et al. / Magnetic Resonance Imaging 33 (2015) 611–617 Table 1 Tissue composition for each voxel location averaged across all patients after segmentation.

613

was performed using the computational cluster at the University of Virginia (http://www.uvacse.virginia.edu/itc-clusters).

Brain region

CSF

Gray matter

White matter

Gray to white matter ratio

2.7. Normalizing GABA values

Frontal Occipital Lateral temporal Basal ganglia Left hippocampus Right hippocampus

0.4% 0.8% 2.2% 1.7% 2.4% 2.5%

25.1% 59.9% 46.9% 57.5% 41.2% 39.3%

74.5% 39.3% 51.2% 40.8% 56.5% 58.2%

0.338 1.526 0.917 1.410 0.729 0.675

Given that NAA is a useful measure of neuronal integrity, we chose to reference GABA to NAA. Referencing to NAA also has the advantage that NAA can be estimated directly from the difference spectra along with GABA and MM09, and does not require a separate water unsuppressed acquisition. The GABA and NAA concentrations may not change proportionally with varying amounts of gray matter and white matter. Hence, tissue fraction differences in different locations may mask the true regional heterogeneity of the GABA/NAA ratio. The GABA/NAA ratio can be expressed as a function of the gray matter and white matter fractions as follows:

characterized. This approach is more prone to subtraction errors, results in lower SNR and incurs double the scan time. An alternative approach for handling the macromolecular contamination would be to account for the co-edited MM30 peak during the post-processing. The macromolecule peak at 0.9 ppm, MM09, does not overlap with any other metabolite in the difference spectrum. The relative peak intensity of the MM30 and MM09 peaks in the difference spectrum was found to be 0.667 [26]. TARQUIN models the MM09 peak as a Gaussian singlet. This can be used for correcting the macromolecular contributions as follows: GABA ¼ ðGABAþÞ−0:667  MM09

ð1Þ

This is comparable to fitting MM30 and MM09 as a combined peak with a fixed peak amplitude ratio [26]. Due to its homogeneous distribution throughout the parenchyma after correcting for the ratio of grey matter within the voxel, N-acetylaspartate (NAA) was used as an internal concentration reference [27]. The NAA peak is not present in the edit-on spectra due to the effect of the editing pulse at 1.9 ppm, but shows up as an inverted peak in the difference spectra. The signal amplitudes obtained from the TARQUIN fits to the NAA peak were used to calculate the tissue-fraction normalized NAA values as shown below. 2.6. Post processing of imaging data In order to determine the proportion of tissue types present in each voxel, each T1-weighted image was segmented using resources available from the open source Advanced Normalization Tools repository [28]. Specifically, brain extraction was performed using the Atropos tool [29] and N4 bias correction [30] facilitated through the use of the ANTs-based scripts, ‘antsBrainExtraction.sh’ and ‘antsAtroposN4.sh’. ANTs-based registration [31] was used to map the labeled brain data provided by the MICCAI 2012 Grand Challenge and Workshop on Multi-Atlas Labeling to each subject [32]. These data consist of 35 T1-weighted images from the OASIS brain database [33] and their corresponding labelings provided by Neuromorphometrics (Neuromorphetrics, Inc., Somerville, MA) under academic subscription. The original 138 labels parcellating the entire brain were reduced to six comprising the following regions: CSF, cortical gray matter, subcortical gray matter, white matter, brain stem, and cerebellum. A consensus labeling (or segmentation) for each subject was derived from the 35 sets of registered labelings using the multi-atlas label fusion (MALF) algorithm [34]. From these segmentations, the CSF, gray matter, and white matter volume ratio values were tabulated. All processing

GABA GABAGM f GM þ GABAWM f WM ¼ NAA NAAGM f GM þ NAAWM f WM

ð2Þ

where GABAGM & GABAWM are the GABA concentrations in the white matter and gray matter, NAAGM & NAAWM are the NAA concentrations in the white matter and gray matter, while fGM and fWM are fractions of gray matter and white matter in the non-CSF tissue within the voxel. The mean tissue composition varies across different locations, as shown in Table 1. The global averaged NAA concentration has been reported to be 1.04–1.78 times higher in the gray matter than the white matter [35–39]. Most of these studies have used single voxel MRS or magnetic resonance spectroscopic imaging (MRSI) techniques and suffer from the problem of limited whole brain coverage, the difficulty of accessing the thin cortical gray matter, and the proximity of gray matter to cranial lipids which can affect the quantitation of NAA. In a recent study using a non-localized whole-brain NAA (WB-NAA) technique that does not suffer from the above limitations, the global averaged gray matter to white matter NAA concentration ratio was 1.5 [40]. The GABA concentration differences in gray matter and white matter are not as well characterized as that of NAA. The effect of regional tissue-fraction variation on NAA was handled as shown below. The ratio of the NAA peak fit amplitudes in two voxels, with identical coil-loading and receiver gain, but different gray matter fractions, fGM and f’GM can be given by: NAA0 NAAGM f 0 GM þ NAAWM f 0 WM ¼ NAA NAAGM f GM þ NAAWM f WM

ð3Þ

The mean gray matter fraction averaged across all locations was 0.457. Substituting, NAAGM/NAAWM = 15 and f’GM = 0.457, the NAA fit amplitude normalized to that of a voxel with gray matter fraction of 0.457 can be given by: 0

NAA ¼



 1:228 NAA 0:5f GM þ 1

ð4Þ

The GABA peak fit amplitude in each voxel was referenced to the corresponding normalized NAA fit amplitude to obtain the GABA/ NAA’ ratio. This ratio is a linear function of the gray matter fraction

Table 2 Shim results using MEGA-PRESS with CHESS water suppression.

Full width half maximum Number of volunteers FWHM

Frontal lobe

Occipital lobe

Lateral temporal lobe

Basal ganglia

Left hippocampus

Right hippocampus

8 18.7 ± 2.6

10 13.8 ± 0.6

8 22.6 ± 2.4

8 23.6 ± 2.0

11 24.2 ± 2.2

12 24.4 ± 2.8

614

C.R. Durst et al. / Magnetic Resonance Imaging 33 (2015) 611–617

Fig. 1. Representative voxel locations, segmentation results, and spectra for each imaging location.

since the denominator is no longer dependent on the tissue-fraction changes. This linear dependence can be exploited for compensating the tissue-fraction dependence of the GABA/NAA’ ratio. The GABA/ NAA’ values in each location was plotted as a function of the gray matter fraction. Spectral quality was assessed by the presence of spectral artifacts, poor spectral fit, large error in fit residuals, or a significantly distorted baseline. Datasets with poor spectral quality were excluded from the analysis. Linear regression analysis was used to estimate the GABA/NAA’ value and the 95% confidence interval (CI) at fGM = 0.457. The number of samples in each location was used to

find the coverage factor t0.975 from the t-distribution table. The standard deviation, σ, was obtained from the following relationship [41]: σ¼

95% CI t 0:975

ð5Þ

Evaluating the GABA/NAA’ value at fGM = 0.457 (henceforth referred to as the normalized GABA/NAA ratio) effectively compensates the regional tissue-fraction variations by normalizing the

C.R. Durst et al. / Magnetic Resonance Imaging 33 (2015) 611–617

615

Fig. 2. Normalized GABA/NAA Box and Whisker plot.

GABA/NAA ratio in any arbitrary voxel to that of a voxel with the mean tissue composition. Eliminating the effect of tissue-fraction driven variation allows us to assess the true regional variation of the GABA/NAA ratio. 2.8. Statistical methods A one-way ANOVA was performed with Tukey's multiple comparison test to compare the normalized GABA/NAA values in each pair of locations. 2.8.1. Statistical software Graphpad Prism 6 (GraphPad Software, Inc., La Jolla, CA) was used to perform all statistical analyses. 3. Results

3.3. Spectroscopy results Voxels in the occipital and frontal lobes tended to be better shimmed (Table 2) than other locations (Fig. 1). Voxels that were poorly shimmed were not included in the analysis. The wide distribution of normalized GABA/NAA values after correction for tissue fraction variation and the contribution of macromolecules is shown in Fig. 2 and tabulated in Table 3. When analyzed by way of a one-way ANOVA performed with Tukey's multiple comparison test, the normalized GABA/NAA ratios varied significantly between brain regions (p b 0.001). Pairwise comparisons of the corrected normalized GABA/NAA ratios show statistically significant variation between the frontal lobe and the basal ganglia, frontal and lateral temporal lobes, and frontal lobe and right hippocampus (Table 4). Variations in the normalized GABA/ NAA ratios trend toward significance between the frontal lobe and left hippocampus, occipital lobe and the frontal lobe, occipital lobe and basal ganglia, and occipital lobe and right hippocampus.

3.1. Subjects Eight male volunteers were enrolled in the study. Ages ranged from 28 to 40 years with an average of 33 and a standard deviation of 4.3. All of the volunteers were healthy. No volunteer reported a history in which abnormal GABA or NAA levels may be implicated, such as epilepsy, neuropsychiatric disorders (including depression, anxiety, or schizophrenia), complex regional pain syndrome, etc. 3.2. Segmentation results

4. Discussion With a few exceptions, prior studies have evaluated the concentration of GABA in a single region of the brain. These studies have identified the relationship between GABA and neuropsychiatric disorders [18,42], effects of medications on GABA concentration [43,44], and lateralization of seizure foci [19,45], for example. Table 4 Pairwise comparisons of the geometric means for normalized GABA/NAA.

The proportion of gray matter within the voxel was greatest for the occipital and lateral temporal lobes (Table 1). The breakdown of components within the left and right hippocampi was similar on average across all patients.

Table 3 Linear regression analysis for normalized GABA/NAA. Location

Number of volunteers

Mean

SD

±95% CI

Frontal Occipital Lateral temporal Basal ganglia Left hippocampus Right hippocampus

7 8 7 7 7 8

0.105 0.152 0.172 0.200 0.157 0.198

0.035 0.042 0.028 0.052 0.018 0.041

0.082 0.097 0.066 0.124 0.044 0.094

Tukey's multiple comparisons test

Mean diff.

95% CI of diff.

Basal ganglia vs. occipital Basal ganglia vs. lateral temporal Basal ganglia vs. left hippocampus Basal ganglia vs. right hippocampus Basal ganglia vs. frontal Occipital vs. lateral temporal Occipital vs. left hippocampus Occipital vs. right hippocampus Occipital vs. frontal Lateral temporal vs. left hippocampus Lateral temporal vs. right hippocampus Lateral temporal vs. frontal Left hippocampus vs. right hippocampus Left hippocampus vs. frontal Right hippocampus vs. frontal

0.048 0.028 0.044 0.002 0.095 −0.020 −0.004 −0.046 0.047 0.015 −0.026 0.067 −0.041 0.051 0.093

−0.011 −0.032 −0.017 −0.056 0.034 −0.079 −0.063 −0.103 −0.012 −0.045 −0.085 0.006 −0.100 −0.009 0.034

to to to to to to to to to to to to to to to

0.107 0.089 0.104 0.061 0.156 0.039 0.055 0.011 0.106 0.076 0.033 0.127 0.017 0.112 0.151

616

C.R. Durst et al. / Magnetic Resonance Imaging 33 (2015) 611–617

However, it has yet to be proven that measurement of GABA concentration in a single region of the brain is a true representation of regional GABA levels. Here, we have found that GABA levels vary significantly throughout the brain. After correcting for tissue fraction variation and macromolecule contribution to the signal, the normalized GABA/NAA ratios vary significantly between regions of the brain (p b 0.001). Our results confirm the findings of recent spectroscopy studies that have suggested a measurable difference in GABA concentrations between two locations within the brain [9,12,15,46]. In a study comparing the medial prefrontal and occipital cortices, the GABA/Cr ratio, corrected for tissue fraction variation, in the occipital lobe was significantly greater than the frontal lobe [47]. In our study, the normalized GABA/ NAA ratio was 45% greater in the occipital lobe than the frontal lobe. Ex-vivo studies have shown that GABA receptor concentrations can vary widely between different regions of the brain [48]. This is in part due to variations in tissue composition and receptor type distribution [49]. These variations represent the varied role of GABA as the primary inhibitory neurotransmitter. These findings coupled with our results suggest that localized measurements of GABA may not be predictive of the concentration of GABA elsewhere in the brain. The methodology employed in this study has several limitations. First, the low number of volunteers limits the power of our findings. Second, any quantification of GABA requires the use of a large voxel. This occasionally makes positioning of the voxel away from the skull base or CSF difficult. Third, the use of a GABA/NAA ratio rather than an absolute quantification of GABA concentration could potentially limit the interpretability of our results. However, efforts were made to minimize the effect of regional tissue-fraction variation driven changes in the NAA level. 5. Conclusion Our study indicates that GABA levels can vary significantly throughout the parenchyma, and that measurement of GABA at one location in the brain may not be indicative of GABA concentrations elsewhere. Based on these findings, we support the recommendation that when studying GABA response, care must be taken when determining the region of interest. References [1] Whittington MA, Traub RD, Jefferys JG. Synchronized oscillations in interneuron networks driven by metabotropic glutamate receptor activation. Nature 1995; 373(6515):612–5. http://dx.doi.org/10.1038/373612a0 [Epub 1995/02/16. PubMed PMID: 7854418]. [2] Petroff OA, Pleban LA, Spencer DD. Symbiosis between in vivo and in vitro NMR spectroscopy: the creatine, N-acetylaspartate, glutamate, and GABA content of the epileptic human brain. Magn Reson Imaging 1995;13(8):1197–211 [Epub 1995/01/01. doi: 0730-725X(95)02033-P [pii]. PubMed PMID: 8750337]. [3] Simister RJ, McLean MA, Barker GJ, Duncan JS. Proton MR spectroscopy of metabolite concentrations in temporal lobe epilepsy and effect of temporal lobe resection. Epilepsy Res 2009;83(2–3):168–76. http://dx.doi.org/10.1016/j. eplepsyres.2008.11.006S0920-1211(08)00319-7 [Epub 2009/01/03. [pii]. PubMed PMID: 19118980]. [4] Bielau H, Steiner J, Mawrin C, Trubner K, Brisch R, Meyer-Lotz G, et al. Dysregulation of GABAergic neurotransmission in mood disorders: a postmortem study. Ann N Y Acad Sci 2007;1096:157–69. http://dx.doi.org/10.1196/ annals.1397.081 [Epub 2007/04/05. PubMed PMID: 17405927]. [5] Walter M, Henning A, Grimm S, Schulte RF, Beck J, Dydak U, et al. The relationship between aberrant neuronal activation in the pregenual anterior cingulate, altered glutamatergic metabolism, and anhedonia in major depression. Arch Gen Psychiatry 2009;66(5):478–86. http://dx.doi.org/10.1001/archgenpsychiatry. 2009.39 [Epub 2009/05/06. PubMed PMID: 19414707]. [6] Farzan F, Barr MS, Levinson AJ, Chen R, Wong W, Fitzgerald PB, et al. Evidence for gamma inhibition deficits in the dorsolateral prefrontal cortex of patients with schizophrenia. Brain 2010;133(Pt 5):1505–14. http://dx.doi.org/10.1093/brain/ awq046 [Epub 2010/03/31. PubMed PMID: 20350936]. [7] Bhat R, Axtell R, Mitra A, Miranda M, Lock C, Tsien RW, et al. Inhibitory role for GABA in autoimmune inflammation. Proc Natl Acad Sci U S A 2010;107(6): 2580–5. http://dx.doi.org/10.1073/pnas.0915139107 [Epub 2010/02/06. PubMed PMID: 20133656; PubMed Central PMCID: PMC2823917].

[8] Boy F, Evans CJ, Edden RA, Singh KD, Husain M, Sumner P. Individual differences in subconscious motor control predicted by GABA concentration in SMA. Curr Biol 2010;20(19):1779–85. http://dx.doi.org/10.1016/j.cub.2010.09.003 [Epub 2010/10/05. PubMed PMID: 20888227; PubMed Central PMCID: PMC3128986]. [9] Floyer-Lea A, Wylezinska M, Kincses T, Matthews PM. Rapid modulation of GABA concentration in human sensorimotor cortex during motor learning. J Neurophysiol 2006;95(3):1639–44. http://dx.doi.org/10.1152/jn.00346.2005 [Epub 2005/10/14. PubMed PMID: 16221751]. [10] Donahue MJ, Near J, Blicher JU, Jezzard P. Baseline GABA concentration and fMRI response. NeuroImage 2010;53(2):392–8. http://dx.doi.org/10.1016/j.neuroimage.2010.07.017 [Epub 2010/07/17. PubMed PMID: 20633664]. [11] Northoff G, Walter M, Schulte RF, Beck J, Dydak U, Henning A, et al. GABA concentrations in the human anterior cingulate cortex predict negative BOLD responses in fMRI. Nat Neurosci 2007;10(12):1515–7. http://dx.doi.org/10. 1038/nn2001 [Epub 2007/11/06. PubMed PMID: 17982452]. [12] Evans CJ, McGonigle DJ, Edden RA. Diurnal stability of gamma-aminobutyric acid concentration in visual and sensorimotor cortex. J Magn Reson Imaging 2010; 31(1):204–9. http://dx.doi.org/10.1002/jmri.21996 [Epub 2009/12/23. PubMed PMID: 20027589]. [13] Sumner P, Edden RA, Bompas A, Evans CJ, Singh KD. More GABA, less distraction: a neurochemical predictor of motor decision speed. Nat Neurosci 2010;13(7): 825–7. http://dx.doi.org/10.1038/nn.2559nn.2559 [Epub 2010/06/01. [pii]. PubMed PMID: 20512136]. [14] O'Gorman RL, Michels L, Edden RA, Murdoch JB, Martin E. In vivo detection of GABA and glutamate with MEGA-PRESS: reproducibility and gender effects. J Magn Reson Imaging 2011;33(5):1262–7. http://dx.doi.org/10.1002/jmri.22520 [Epub 2011/04/22. PubMed PMID: 21509888; PubMed Central PMCID: PMC3154619]. [15] Gao F, Edden RA, Li M, Puts NA, Wang G, Liu C, et al. Edited magnetic resonance spectroscopy detects an age-related decline in brain GABA levels. NeuroImage 2013;78:75–82. http://dx.doi.org/10.1016/j.neuroimage.2013.04.012 [Epub 2013/04/17. PubMed PMID: 23587685; PubMed Central PMCID: PMC3716005]. [16] Mescher M, Tannus A, Johnson M, Garwood M. Solvent suppression using selective echo dephasing. J Magn Reson Ser A 1996;123(2):226–9. [17] Mescher M, Merkle H, Kirsch J, Garwood M, Gruetter R. Simultaneous in vivo spectral editing and water suppression. NMR Biomed 1998;11(6):266–72. http://dx.doi.org/10.1002/(SICI)1099-1492(199810)11:6b266::AIDNBM530N3.0.CO;2-J [Epub 1998/11/05. [pii]. PubMed PMID: 9802468]. [18] Petroff OA, Rothman DL, Behar KL, Mattson RH. Low brain GABA level is associated with poor seizure control. Ann Neurol 1996;40(6):908–11. http://dx. doi.org/10.1002/ana.410400613 [PubMed PMID: 9007096]. [19] Cendes F, Caramanos Z, Andermann F, Dubeau F, Arnold DL. Proton magnetic resonance spectroscopic imaging and magnetic resonance imaging volumetry in the lateralization of temporal lobe epilepsy: a series of 100 patients. Ann Neurol 1997;42(5):737–46. http://dx.doi.org/10.1002/ana.410420510 [PubMed PMID: 9392573]. [20] Hammen T, Stefan H, Eberhardt K, W‐Huk B, Tomandl B. Clinical applications of 1H‐MR spectroscopy in the evaluation of epilepsies–what do pathological spectra stand for with regard to current results and what answers do they give to common clinical questions concerning the treatment of epilepsies? Acta Neurol Scand 2003;108(4):223–38. [21] Riederer F, Bittsansky M, Schmidt C, Mlynarik V, Baumgartner C, Moser E, et al. 1H magnetic resonance spectroscopy at 3 T in cryptogenic and mesial temporal lobe epilepsy. NMR Biomed 2006;19(5):544–53. http://dx.doi.org/10.1002/nbm. 1029 [PubMed PMID: 16521092]. [22] Wilson M, Reynolds G, Kauppinen RA, Arvanitis TN, Peet AC. A constrained leastsquares approach to the automated quantitation of in vivo (1)H magnetic resonance spectroscopy data. Magn Reson Med 2011;65(1):1–12. http://dx.doi. org/10.1002/mrm.22579 [Epub 2010/09/30. PubMed PMID: 20878762]. [23] Henry PG, Dautry C, Hantraye P, Bloch G. Brain GABA editing without macromolecule contamination. Magn Reson Med 2001;45(3):517–20 [PubMed PMID: 11241712]. [24] Edden RA, Puts NA, Barker PB. Macromolecule-suppressed GABA-edited magnetic resonance spectroscopy at 3 T. Magn Reson Med 2012;68(3): 657–61. http://dx.doi.org/10.1002/mrm.24391 [PubMed PMID: 22777748; PubMed Central PMCID: PMC3459680]. [25] Rothman DL, Petroff OA, Behar KL, Mattson RH. Localized 1H NMR measurements of gamma-aminobutyric acid in human brain in vivo. Proc Natl Acad Sci U S A 1993;90(12):5662–6 [PubMed PMID: 8516315; PubMed Central PMCID: PMC46781]. [26] Long ZM JB, Xu J, Dydak U, editors. GABA fitting for MEGA-PRESS sequences with different selective inversion frequenciesProceedings International Society of Magnetic Resonance Medicine; 2011. [27] Pouwels PJ, Frahm J. Regional metabolite concentrations in human brain as determined by quantitative localized proton MRS. Magn Reson Med 1998;39(1): 53–60 [PubMed PMID: 9438437]. [28] Avants BB, Tustison N, Song G. Advanced Normalization Tools (ANTS). Insight J 2009;2:1–35. [29] Avants BB, Tustison NJ, Wu J, Cook PA, Gee JC. An open source multivariate framework for n-tissue segmentation with evaluation on public data. Neuroinformatics 2011;9(4):381–400. [30] Tustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA, et al. N4ITK: improved N3 bias correction. IEEE Trans Med Imaging 2010;29(6):1310–20. [31] Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration.

C.R. Durst et al. / Magnetic Resonance Imaging 33 (2015) 611–617

[32]

[33]

[34]

[35]

[36] [37]

[38]

[39]

[40]

NeuroImage 2011;54(3):2033–44. http://dx.doi.org/10.1016/j.neuroimage. 2010.09.025 [Epub 2010/09/21. PubMed PMID: 20851191; PubMed Central PMCID: PMC3065962]. Landman B, Warfield S, editors. MICCAI 2012 workshop on multi-atlas labelingMedical Image Computing and Computer Assisted Intervention Conference 2012: MICCAI 2012 Grand Challenge and Workshop on Multi-Atlas Labeling Challenge Results; 2012. Marcus DS, Wang TH, Parker J, Csernansky JG, Morris JC, Buckner RL. Open Access Series of Imaging Studies (OASIS): cross-sectional MRI data in young, middle aged, nondemented, and demented older adults. J Cogn Neurosci 2007;19(9): 1498–507. http://dx.doi.org/10.1162/jocn.2007.19.9.1498 [Epub 2007/08/24. PubMed PMID: 17714011]. Wang H, Yushkevich PA. Multi-atlas segmentation with joint label fusion and corrective learning-an open source implementation. Front Neuroinformatics 2013;7:27. http://dx.doi.org/10.3389/fninf.2013.00027 [Epub 2013/12/10. PubMed PMID: 24319427; PubMed Central PMCID: PMC3837555]. Tal A, Kirov II, Grossman RI, Gonen O. The role of gray and white matter segmentation in quantitative proton MR spectroscopic imaging. NMR Biomed 2012;25(12):1392–400. http://dx.doi.org/10.1002/nbm.2812 [PubMed PMID: 22714729; PubMed Central PMCID: PMC3449040]. Kreis R. Quantitative localized b sup N 1b/sup N H MR spectroscopy for clinical use. Prog Nucl Magn Reson Spectrosc 1997;31(2):155–95. Woermann FG, McLean MA, Bartlett PA, Barker GJ, Duncan JS. Quantitative short echo time proton magnetic resonance spectroscopic imaging study of malformations of cortical development causing epilepsy. Brain 2001;124(Pt 2): 427–36 [PubMed PMID: 11157569]. McLean MA, Barker GJ. Concentrations and magnetization transfer ratios of metabolites in gray and white matter. Magn Reson Med 2006;56(6):1365–70. http://dx.doi.org/10.1002/mrm.21070 [PubMed PMID: 17051529]. Narayana P, Fotedar L, Jackson E, Bohan T, Butler I, Wolinsky J. Regional b i N in vivob/i N proton magnetic resonance spectroscopy of brain. J Magn Reson 1989; 83(1):44–52. Inglese M, Rusinek H, George IC, Babb JS, Grossman RI, Gonen O. Global average gray and white matter N-acetylaspartate concentration in the human brain.

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

617

NeuroImage 2008;41(2):270–6. http://dx.doi.org/10.1016/j.neuroimage.2008. 02.034 [PubMed PMID: 18400521; PubMed Central PMCID: PMC2486451]. Croarkin C, Tobias P. NIST/SEMATECH e-handbook of statistical methods. NIST/ SEMATECH. July Available online: http://www itl nist gov/div898/handbook; 2006. Kegeles LS, Shungu DC, Anjilvel S, Chan S, Ellis SP, Xanthopoulos E, et al. Hippocampal pathology in schizophrenia: magnetic resonance imaging and spectroscopy studies. Psychiatry Res 2000;98(3):163–75 [PubMed PMID: 10821999]. Petroff OA, Rothman DL, Behar KL, Lamoureux D, Mattson RH. The effect of gabapentin on brain gamma-aminobutyric acid in patients with epilepsy. Ann Neurol 1996;39(1):95–9. http://dx.doi.org/10.1002/ana.410390114 [PubMed PMID: 8572673]. Petroff OA, Rothman DL, Behar KL, Mattson RH. Human brain GABA levels rise after initiation of vigabatrin therapy but fail to rise further with increasing dose. Neurology 1996;46(5):1459–63 [PubMed PMID: 8628502]. Doelken MT, Richter G, Stefan H, Doerfler A, Noemayr A, Kuwert T, et al. Multimodal coregistration in patients with temporal lobe epilepsy–results of different imaging modalities in lateralization of the affected hemisphere in MR imaging positive and negative subgroups. AJNR Am J Neuroradiol 2007;28(3): 449–54 [PubMed PMID: 17353311]. Puts NA, Edden RA. In vivo magnetic resonance spectroscopy of GABA: a methodological review. Prog Nucl Magn Reson Spectrosc 2012;60:29–41. http:// dx.doi.org/10.1016/j.pnmrs.2011.06.001S0079-6565(11)00043-4 [Epub 2012/ 02/02. [pii]. PubMed PMID: 22293397; PubMed Central PMCID: PMC3383792]. van der Veen JW, Shen J. Regional difference in GABA levels between medial prefrontal and occipital cortices. J Magn Reson Imaging 2013;38(3):745–50. http://dx.doi.org/10.1002/jmri.24009 [PubMed PMID: 23349060; PubMed Central PMCID: PMC3638064]. Palacios JM, Wamsley JK, Kuhar MJ. High affinity GABA receptors-autoradiographic localization. Brain Res 1981;222(2):285–307 [Epub 1981/10/19. doi: 0006-8993(81)91034-9 [pii]. PubMed PMID: 6269695]. Sieghart W, Sperk G. Subunit composition, distribution and function of GABA(A) receptor subtypes. Curr Top Med Chem 2002;2(8):795–816 [PubMed PMID: 12171572].